sdlc-execute

Implements an approved plan by dispatching one subagent per task in plan order.

Updated Sep 11, 2026
One-click install
npx skills add https://github.com/zhixuan312/zz-stack --skill sdlc-execute-zhixuan312
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: sdlc-execute
Source: https://github.com/zhixuan312/zz-stack/tree/main/catalog/sdlc/sdlc-flow/skills/sdlc-execute
Command: npx skills add https://github.com/zhixuan312/zz-stack --skill sdlc-execute-zhixuan312

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Turning an approved implementation plan into working code without losing accountability is hard: workers drift from contracts, checks get weakened, and partial runs get reported as done. This Skill orchestrates plan execution so every task's contract is made true, every plan-authored check actually runs, and what changed is reported from the tree rather than from worker claims. ## Core Features & Use Cases - Sequential subagent dispatch: One worker per task in plan order, each handed the task's contract verbatim, with the orchestrator keeping the sequence and the branch. - Check freeze and activation: Every plan-authored check is frozen up front outside the target's discovery path, activated one at a time immediately before its task, and byte-compared afterwards so no worker can weaken it. - Full-suite gate after every task: Per-task checks prove the task did its job; the gate proves the rest of the project still works, and a red gate stops the sequence. - Use Case: A team has an audited, approved plan.md with five tasks and declared acceptance checks. This Skill cuts a branch, freezes the checks, dispatches five workers one at a time, runs checks and the gate after each, and reports the real diff before asking whether to commit. ## Quick Start Execute the approved plan.md in this repository by dispatching one subagent per task in order and running each task's checks plus the full-suite gate.

Frequently Asked Questions about sdlc-execute

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I execute an approved implementation plan with subagents?▼

Verify plan.md is approved via its frontmatter, cut a task branch, freeze every plan-authored check, then dispatch one subagent per task in plan order. After each task, run its checks and the full-suite gate before starting the next.

What happens when a task fails during plan execution?▼

Re-dispatch only the failed task with a fresh worker told what the previous attempt tried. If two workers fail the same task the same way, the contract is probably wrong, so take it back to the planning stage rather than retrying a third time.

Can I use this without a git repository?▼

Yes. On a non-git target, edits happen in place with no commit, and that is stated plainly rather than inventing a repository. The orchestration, check freezing, and gate runs still apply.

Why is a check that could not run not treated as failed?▼

A check that dies on a denied port bind, missing binary, sandbox restriction, or timeout never executed, so it produced no verdict. The task is judged on its contract and the checks that actually ran, and unverifiable checks are named so someone can run them elsewhere.

When should a platform tool problem be reported instead of worked around?▼

When a tool refuses what its own description says it accepts or breaks reproducibly, file it with bug_report naming the tool and the interrupted initiative, then continue the task. A task that simply failed is not a platform bug and is reported as a plan record instead.